We recently tested serum from a human immunodeficiency virus-infected patient for the presence of cryptococcal antigen using the Meridian latex agglutination (LA) test (Cryptococcal Antigen Latex Agglutination System). Two pronase-treated serum specimens from the patient had LA titers of 80 and 160, but the patient had no evidence of cryptococcal disease. The serum was negative for rheumatoid factor, a well-documented cause of false-positive LA reactions. Seven blood culture supernatants from the patient were also LA positive, but were culture negative for cryptococcus. When the sera and blood culture supernatants were treated with 0.01 M 2-beta-mercaptoethanol (2-ME), the agglutinating activity was ablated. Similar results were seen when the sera were tested by two other commercial LA assays. Serum and cerebrospinal fluid specimens from patients with confirmed cryptococcal disease were treated with 2-ME, and the results were compared with those obtained after pronase (sera) or heat (cerebrospinal fluid) inactivation. The titers were identical (n = 56) or within 1 dilution (n = 3). One hundred serum specimens from human immunodeficiency virus-seropositive patients with no known history of cryptococcal disease were examined to determine the frequency of false-positive reactivity in this patient population. Of this group, three were positive following pronase treatment. One remained positive after 2-ME treatment; the remaining two were negative. These data indicate that 2-ME can be used to eliminate nonspecific reactivity in the LA test without affecting true-positive results.